## Inspiration DysphagiaSense is personal. One teammate saw a family member struggle with swallowing issues that were hard to detect early. Current screening is expensive and clinical, so many people miss early warning signs. We wanted a simple, accessible way to catch problems sooner and support families in everyday settings.

## What it does DysphagiaSense records a short swallow sound and analyzes it to detect potential swallowing issues. It extracts acoustic features and uses a transparent rule-based system to classify risk as low, medium, or high, then provides clear, easy-to-understand feedback with next steps.

## How we built it We built a full-stack system with a React frontend and FastAPI backend. Audio is captured using the browser microphone, processed with libraries like librosa, and analyzed using a rule-based classifier. AI (Gemini) generates simple explanations, while charts and visuals show the results clearly.

## Challenges we ran into Handling different audio formats across devices, especially iPhones, was difficult. Ensuring reliable feature extraction from noisy recordings was also challenging. We also had to design a system that stays explainable while still being useful without clinical data.

## Accomplishments that we're proud of We built a working end-to-end product in a short time. The system provides real-time analysis, clear visualizations, and understandable feedback. Most importantly, we created a tool that can make early screening more accessible to people who need it.

## What we learned We learned how to process and analyze audio signals, design explainable AI systems, and build full-stack applications quickly. We also gained insight into healthcare challenges and the importance of usability and clarity in sensitive applications.

## What's next for DysphagiaSense Next steps include clinical validation with experts, improving accuracy with machine learning, adding long-term tracking, and building a mobile app. The goal is to move from a prototype to a reliable, widely accessible screening tool.

Built With

  • fastapi-(uvicorn)
  • google-gemini-2.5-flash
  • librosa-+-numpy
  • mediarecorder-&-web-audio-api
  • pydub-+-ffmpeg
  • react-(typescript
  • recharts
  • rule-based-classifier
  • vite)
Share this project:

Updates